Skip to contentClass figures are pooled from every eligible student on the same test.
Question health
Every question gets one label, and the reasons for it.
Open the diagnostic columns on a test's question table to see the label for each question. Click a question below to see why it got its label and which measures were used.
Tap a question to see why it got its label.
Y10 Geography · Urban issuesExample data · 26 students
Rules, checked in this order. The first one that fits sets the label.
- 1Not compared. An optional choice question, or fewer than 5 students counted.
- 2Too easy · ceiling. An average of 95% or more, or everyone on full marks.
- 3Review. The AI markers agreed exactly on under 60% of scripts, 30% or more were flagged for review, or you changed 30% or more of the marks.
- 4By average score. Under 40%: Fine · reteach. 40 to 59%: Fine · retrieval. 60 to 84%: Strong. 85% or more: Useful.
Example data. Rules run in order: not comparable, too easy, marking concerns, then the average score.
ReviewQ5 · 4 marks
Explain how informal settlements develop.Why it got this label
- The two AI markers agreed exactly on only 52% of scripts.
- 38% of answers were flagged for review.
- Teachers changed the AI mark on 35% of scripts, mostly down, so the AI was too generous. Worth a look at the mark scheme.
The measures behind it
Example data. "Separates students" compares each student's mark on this question with their total. On one class it is always shown as provisional, and it never sets the label.- Average score
- 55%
- AI markers agreed exactly
- 52%
- Flagged for review
- 38%
- Marks changed by teachers
- 9 scripts, all down
- Attempted
- 26 of 26
- Legibility
- 82%
- Separates students
- 0.21, provisional (range -0.19 to 0.55)
Writing insights
How each student writes, from answers you have already marked.
Four writing measures per student, compared with the class on the same test. Writing flags show what is getting in the way, with the answers they came from.
Measures describe the writing. They are not an attainment score, and sentence or vocabulary measures never raise a flag on their own.
How Aisha is writing · Example dataUnderdeveloped answers · PersistentStructure and development · Repeated
- 1Four writing measuresSentence length, vocabulary range, common words, words per mark.
- 2Writing flags with evidenceEach links to the answers it was seen in.
The student profile works best on a larger screen.
Writing insight · advisoryHow Aisha is writing
Example data · Y10 GeographyAverage sentence14.2 wordsWords in each sentence−1.9 words vs class (16.1 words)
Vocabulary range0.71Different words in each 50-word stretch+0.02 vs class (0.69)
Common-word coverage92%Words found in a common-word list+1 point vs class (91%)
Words per awarded mark8.4 wordsProse words divided by marks awarded−1.2 words vs class (9.6 words)
Writing flags on this test
Possible hidden proficiencyShort answers averaged 86% but extended answers 38%. Aisha may know more than her longer writing shows.
Written summary · made when you ask · exampleAccurate short answers. Longer answers stop after the first point.An AI-assisted summary of the student's own marked answers. Review it before acting on it.
Patterns over time
One slip is seen. A habit is persistent.
Each flag is tracked across a student's marked tests in the subject. Tap the tests where the flag appeared and watch its status change.
Aisha · Underdeveloped answersExample data · last 5 tests
StatusPersistentTrendWorsening
- PersistentIn 3 or more of the last 5 tests.
- RepeatedIn 2 or more of the last 3 tests.
- SeenAny other time it appears, including once.
Tap a test. Persistent: 3 of the last 5. Repeated: 2 of the last 3. Seen: any other time.
Trend compares the latest 3 tests with the ones before, and only shows once there are at least 4 marked tests.Class view · results pageWriting insights for 10GFrom this class's marked answers. No extra AI.
Top writing barriers
Underdeveloped answers7 students2 high · 3 medium · 2 low
Command-word mismatch5 students1 high · 4 medium
Structure and development4 students3 medium · 1 low
Students needing follow-up
Example data. Students with the most, and most severe, flags come first.Aisha PatelUnderdeveloped answers · Structure and development
Omar KhanCommand-word mismatch · Handwriting legibility
Examiner comments
Got the examiner's report? Add it.
Type comments against a question or upload the report, then check any unsure matches. Comments are advice only: they never add or remove credit.
Examiner comments sit next to your mark scheme for each question. They help the markers read your scheme, and they appear on each question's detail page in analysis.
Type them in, or upload the reportAdd notes against any question, or upload the report and the comments are matched to your questions.
Check the unsure matchesWhen a match is low confidence, it is listed for you to check before marking.
Advice onlyComments never add or remove credit. Your mark scheme decides the mark.
What you control
Signals for you to read, not verdicts.
Out of the way until you want itDiagnostic columns stay hidden until you open them, and you choose which ones to show: label, separates students, marker reliability, attempts and evidence quality.Diagnostic columns stay hidden until you open them.
Honest about one classWith fewer than 30 students, the page says the comparisons are indicative only. Handwriting that was hard to read is shown as a caution, never blamed on the question.
No extra AI for the numbersQuestion health, writing measures and writing flags are counted by fixed rules from the marking you already have. The same answers always give the same result.Counted by fixed rules from marking you already have.
The summary is yours to checkThe written summary is AI-assisted and only made when you ask. If marks change afterwards, it is labelled "May be out of date" until you regenerate it. Writing insights come with AI-marked tests, not tests you mark by hand.AI-assisted, made only when you ask, for you to review.
Fair to every studentLanguage measures are not collected for students recorded as EAL. Very short answers, and scanned answers that were hard to read, are left out of the measures.No language measures for students recorded as EAL.
Teacher review comes firstTeachers should review writing insights alongside the original answers before acting on them.
Start with this week's test.
Use the paper you already set. You stay in control of every mark, your students get better feedback, and you see what to teach next.
No subscription required. You approve every mark.Getting started
- 1Upload your test and mark schemeThe questions and mark-scheme points are pulled out for you to check.
- 2Add the class's scriptsScan the pile as one PDF, upload a file per student, or let students scan a QR code on their devices.
- 3Review and releaseEach student starts marking as soon as their name is confirmed. You check the flagged answers and release.